Experiments in AI art
Anglo-Saxon warrior queens and spider witches created in Midjourney AI
When I first came across AI art generators, they were exploding in popularity on social media, in the summer of 2022. I was immediately excited by the possibilities in how I could use it, as my project focused on compositing images inspired by myths. I could introduce assets of a fictional nature, to add to scenes to create narrative depictions. I experimented with some characters at first, and found that AI had some inherent biases that I would have to counter with tweaking the prompts. I used MidJourney for the images posted here. Inspired by myth and history, I started by asking it to create spider witches and warrior queens.
Surreal gothic fantasy scenes generated from one of my photographs as an image prompt, using Midjourney
Another issue that seemed to be coming up a lot in discussions on social media, plenty of them heated, was the question of authorship. This took on additional significance once it was revealed that artists work had been used to train the language models but without their permission. It would be very easy to copy the style of their work without attribution. While I still feel that AI art programs are an exciting development, I would only feel okay using them if I prompted them to generate images from my own source material, which become a new development. I tried creating some scenes from an image I took, but the AI may still have used other styles without me specifying anyone.
My current thoughts are undecided, and I’m keeping an eye on developments. I do think that mostly, images produced by AI can be original, and it’s not simply a case of copying others. The quality and originality of AI art is ultimately down to the user, and how much they hone their ability to craft prompts, and if they have original ideas to start with. For now I’m happy with my current photographic style, which has developed throughout this project, as you will see in the upcoming posts.
